Amazon Web Services is using AI to change how its pricing team works, not just how fast it works.
John Felton, SVP and CFO of Amazon Web Services, said the team converted an 18-tab Excel model into a chatbot interface that lets employees evaluate customer deals through natural-language questions. The model can test what happens if a price falls 20%, how payment terms affect a deal, or where the break-even point lies. Employees who still prefer spreadsheets can export the model back to Excel.
Felton said the point was not merely convenience. It changed how employees interact with the underlying analysis.
He also described another internal tool that compares payment terms in customer contracts with those recorded in AWS’s payment system. Before the AI agent, employees checked only a sample of contracts. Now the team can check all of them.
'Let’s not just do a sample size, let’s do all of it,' Felton said.
Felton said the broader shift in AI adoption is moving away from improving existing work and toward rethinking what the business can do. Two years ago, customer conversations focused mostly on productivity and cost cutting. Now, he said, there is more focus on new revenue products and new customer experiences.
'How do I use AI for new revenue products, for new customer experiences?' he said.
That matters because AWS is now a $169 billion annualized business, and Felton said AI adoption across companies is still uneven. Many customers are still early in broader cloud and IT modernization, which he described as a prerequisite for deploying AI at scale.
Inside AWS’s finance organization, Felton said he wants employees to identify AI opportunities themselves rather than wait for direction from management. He said the company is encouraging experimentation, and he uses Amazon Quick, an AI assistant for work, to query supporting materials for board meetings and find answers in underlying files.
The shift is simple to describe and harder to ignore. AI is no longer only a labor-saving tool. At AWS, it is becoming a management tool for expanding what the business can actually see, test and enforce.
That is the part the market will notice. When a pricing team can examine every contract instead of a sample, and when a model once buried in 18 tabs becomes a conversational interface, the real gain is control. Capital rewards clearer rules, better leverage and faster decisions. In this case, the administrative burden does not disappear, but it gets organized into something more usable for the people making the call.
